ARTICLES
Variegated gene expression caused by cell-specific
long-range DNA interactions
Daan Noordermeer1,4,5 , Elzo de Wit1,5 , Petra Klous1,5 , Harmen van de Werken1 , Marieke Simonis1 ,
Melissa Lopez-Jones2 , Bert Eussen3 , Annelies de Klein3 , Robert H. Singer2 and Wouter de Laat1,6
Mammalian genomes contain numerous regulatory DNA sites with unknown target genes. We used mice with an extra β-globin
locus control region (LCR) to investigate how a regulator searches the genome for target genes. We find that the LCR samples a
restricted nuclear subvolume, wherein it preferentially contacts genes controlled by shared transcription factors. No contacted
gene is detectably upregulated except for endogenous β-globin genes located on another chromosome. This demonstrates
genetically that mammalian trans activation is possible, but suggests that it will be rare. Trans activation occurs not pan-cellularly,
but in ‘jackpot’ cells enriched for the interchromosomal interaction. Therefore, cell-specific long-range DNA contacts can cause
variegated expression.
High-resolution profiling of transcription-factor binding sites, the
discovery of conserved genetic elements and identification of regulatory
sites using technologies such as DNaseI hypersensitive site mapping1 ,
has demonstrated that the number of genomic sites with transcription
regulatory potential far exceeds the number of genes. One of the main
challenges of the post-genomic era therefore is to assign function
to each element. This requires understanding of the capacity of
regulatory sites to reach over distance and identify specific target
genes at the single-cell level. It is known that mammalian DNA
elements can modulate the activity of distant genes on the same
chromosome, up to a genomic distance of over a megabase2,3 . The
three-dimensional structure of the mammalian genome facilitates
long-range gene regulation. This was first shown for the β-globin
locus, which contains multiple β-globin genes arranged in the order
of their developmental expression (Fig. 1a). Upstream of the genes
resides a large regulatory DNA element that enhances expression of the
β-globin genes up to ∼100-fold4 . The element is called a locus control
region (LCR), as it has the capacity to confer position-independent and
copy-number-dependent expression on reporter genes in transgenic
assays in addition to its classical enhancer activity5 . Otherwise, the
β-globin LCR is not different from classical enhancers: it upregulates
gene expression over distance, it functions in a tissue-specific manner
and genes compete for its activity6,7 . At its endogenous location, the
LCR enhances expression of the β-globin genes through physical
interactions, thereby looping out the intervening DNA that may carry
non-responding genes8–10 (Supplementary Fig. S1). A similar mode of
action involving chromatin looping has been shown for enhancers in
several other gene loci11,12 .
At a higher-order level of organization, microscopy studies have
shown that genes can occupy differential positions in the nucleus
depending on their expression status13–15 and that regulatory DNA
elements are instrumental in targeting these genomic regions to specific
nuclear positions16–19 . Moreover, recent observations suggest that
functionally related genes on the same and different chromosomes
may come together in the nuclear space for co-transcription20 . Finally,
some reports have suggested functional communication between
regulatory sites located on different chromosomes21–25 . Collectively,
these studies raise questions on how genes and regulatory sequences
explore the nuclear space to engage in functional crosstalk with
preferred genomic partners.
RESULTS
Exploring mammalian transvection
To investigate the ability of regulatory DNA elements to relocate
chromosomal regions in the nuclear interior and search for preferred
target genes, we used transgenic mice with the human erythroid-specific
β-globin LCR (hLCR) site-specifically integrated into a gene-dense
region on chromosome 8. This region, 8C3–C4, contains many
housekeeping genes18 . Two transgenic lines were available, LCR-S and
LCR-AS, each carrying the hLCR without globin genes at 8C3–C4, but
in opposite orientations (Fig. 1b). In a previous study we showed that in
erythroid cells each LCR detectably upregulates several housekeeping
1
Hubrecht Institute-KNAW and University Medical Center Utrecht, Uppsalalaan 8, 3584 CT Utrecht, The Netherlands. 2 Department of Anatomy and Structural Biology,
Albert Einstein College of Medicine, Bronx, New York 10461, USA. 3 Department of Clinical Genetics, Erasmus Medical Centre, PO Box 2040, 3000 CA Rotterdam,
The Netherlands. 4 Present address: Laboratory of Developmental Genomics, School of Life Sciences, Swiss Federal Institute of Technology (EPFL), Lausanne,
Switzerland. 5 These authors contributed equally to this work.
6
Correspondence should be addressed to W.d.L. (e-mail: [email protected])
Received 12 September 2010; accepted 9 May 2011; published online 26 June 2011; DOI: 10.1038/ncb2278
944
NATURE CELL BIOLOGY VOLUME 13 | NUMBER 8 | AUGUST 2011
© 2011 Macmillan Publishers Limited. All rights reserved.
ARTICLES
a
Mouse β-globin locus
(mouse chromosome 7)
LCR β-globin genes
654 3 21 εy βh1 βmaj βmin
–85HS –62/–60HSs
2
5
4
1
E
Human β-globin locus
(human chromosome 11)
6
2
Gγ Aγ
P
5
4
3
1
hLCR-S
1
E/F
5
G
P: Puromycin
N: Neomycin
G: GFP
N
E12.5 fetal liver
0.5
0
– AS S
Aγ-globin: AS AS AS S
e
DNA FISH
Separate:
8C3–C4
(BAC A15)
8C3–C4
(BAC P23)
DAPI
0.5
0
– AS S – AS S
AS AS AS S S S
S S
Co-localized:
1.0
DNA FISH
E12.5 fetal liver
Co-localization
frequency
1.0
Relative
expression
Relative
expression
5 kb
G
N
c E10.5 Embryonic blood
d
20 kb
P
Rad23a
– AS S
3
1
4
3
Rad23a
hLCR:
2
B B
E Embryonic blood
F Fetal liver
B Adult bone marrow
2
2
8C3–C4
Aγ-globin–GFP-AS
Aγ-globin–GFP-S
4
3'HS1
1
β-globin gene
DNaseI hypersensitive site
Olfactory receptor gene
b hLCR-AS
3
5
E
/ HS
2
F/B F/B
LCR β-globin genes
4 3
7
E
54 3 2 1 ε
–110/–107HSs
8
3'HS1
1
3
is generally not observed in mammals28 , but may occur in special
instances where mono-allelic expression needs to be ensured29–31 .
As mammalian chromosomes intermingle extensively32 , individual
chromosomal segments may indeed invade the territories of other,
possibly homologous, chromosomes. We generated transgenic mice
carrying a human Aγ-globin reporter gene in one or the other
orientation at exactly the same location in 8C3–C4 (Fig. 1b). The
Aγ-globin gene is a human fetal globin gene that, in mouse transgenics
carrying a full human globin cluster including the hLCR, is most
highly expressed between embryonic day 10.5 and 12.5 (ref. 33;
Fig. 1a). Crossing the Aγ-globin reporter mice with transgenic mice
carrying the hLCR on the homologue revealed no increase in Aγ-globin
gene expression at any of the relevant developmental stages (Fig. 1c).
Fluorescence in situ hybridization (FISH) experiments also did not
reveal increased interaction between the homologues (Fig. 1d,e). Therefore, our experimental system provides no evidence for mammalian
transvection involving the hLCR at 8C3–C4. Although these data do
not rule out mammalian transvection at other genomic locations, they
do strongly suggest that the ectopic hLCR does not have unlimited
freedom to search the nuclear interior for natural target genes.
NS
NS
6.0
4.0
2.0
0
hLCR:
Aγ-globin:
–
–
S
–
S
S
Figure 1 An ectopic LCR does not activate a natural target gene on the
homologous chromosome. (a) Schematic representation of the endogenous
mouse and human β-globin loci. Below each globin gene, gene activity in
(transgenic) mice is indicated. (b) Targeting strategy for the insertion of
the human β-globin LCR and a human Aγ-globin–green fluorescent protein
(GFP) reporter gene into the 8C3–C4 locus on mouse chromosome eight.
(c) RT–qPCR of Aγ-globin transcript levels, normalized to Hprt1 transcript
levels. Data are from at least two independent samples. (d) Representative
examples of DNA FISH showing co-localized and separate 8C3–C4 alleles.
DAPI, 4,6-diamidino-2-phenylindole. Scale bar: 2 µm. (e) Co-localization
frequencies of 8C3–C4 alleles. Significance levels are indicated above the
graph (G -test).
genes that directly surround the integration site up to sixfold, with
the most distal one being 150 kilobases (kb) away. Microscopy studies
revealed that both LCRs positioned 8C3–C4 more often outside its
own chromosome territory, raising the question of to where the LCR
migrates its integration site.
We reasoned that, in the most extreme situation, the hLCR could
‘search’ for a natural target gene present anywhere in the genome,
including on its homologous chromosome, as is seen in transvection.
The term transvection was coined to describe transcriptional regulation
across (paired) homologous chromosomes, a phenomenon mainly
studied in Drosophila, where homologues frequently pair26,27 . Pairing
LCR motion is limited by chromosomal context
To investigate in more detail the ability of the hLCR to actively determine its genomic environment, we applied chromosome conformation
capture34 on chip (4C) technology to E14.5 fetal livers of wild-type and
homozygous LCR-AS mice (Supplementary Fig. S2). 4C captures and
identifies spatially proximal DNA sequences to enable an unbiased scan
for DNA elements interacting with a locus of choice35 . Analysis of the
4C data revealed that the ratio of inter- over intrachromosomal captures
increased in the LCR-AS mice (Supplementary Fig. S3), in agreement
with the hLCR causing looping out from the chromosome territory
(ref. 18). This might be a reflection of the LCR actively engaging in
interactions with new interchromosomal partners. To identify specific
DNA interactions, sliding-window algorithms were applied that scan
the linear chromosome templates for significant clustering of independently captured sequences36,37 . Using both conventional analysis and a
newly developed high-resolution analysis, we identified a highly similar
set of interacting regions for 8C3–C4 with or without an integrated
hLCR, both on the same chromosome (in cis) and on different
chromosomes (in trans) (Fig. 2a–c and Supplementary Figs S4, S5a). Extensive fluorescence in situ hybridization experiments on cryo-sections
(cryo-FISH; ref. 32) validated the 4C data (Fig. 2d,e). Some regions in
trans showed increased interaction frequencies with the hLCR, again
in agreement with the LCR causing looping out of the chromosome
territory. However, we did not find chromosomal regions that exclusively interacted with the hLCR. Altogether, this demonstrates that the
hLCR at 8C3–C4 does not search the genome to contact new preferred
interaction partners. Rather, the chromosomal context of 8C3–C4
dictates the nuclear space that can be explored by the integrated hLCR.
The LCR contacts GATA-1 and EKLF regulated genes
Whereas the overall genomic environment did not change, a
quantitative comparison of 4C results showed that the hLCR at 8C3–C4
captured a subset of pre-existing interchromosomal interaction
partners more efficiently. This was most obvious for the α-globin
locus on chromosome 11, but also for the endogenous β-globin
NATURE CELL BIOLOGY VOLUME 13 | NUMBER 8 | AUGUST 2011
© 2011 Macmillan Publishers Limited. All rights reserved.
945
ARTICLES
a
8C3–C4
max.
Chromosome 8
hLCR-AS 1
0
max.
hLCR-AS 2
0
max.
Wild type 1
0
max.
Wild type 2
0
0
20
80
60
Chromosomal location (Mb)
40
b
100
120
8C3–C4
200
hLCR-AS
0
200
10
100
0
8
0
20
40
60
80
Chromosomal location (Mb)
β-globin
c
200
hLCR-AS
120
6
α-globin
Chromosome 7
Chromosome 11
4
100
Window size
Wild type
100
0
200
–log10 P-value
Window size
Wild type
Chromosome 8
100
2
100
0
0
0
50
100
Chromosomal location (Mb)
150
e
d Cryo-FISH
50
Chromosomal location (Mb)
100
Cryo-FISH
Co-localization frequency (%)
Co-localized:
Separate:
8C3–C4
∗
8
∗
∗
∗
Fetal liver hLCR-AS
6
Fetal liver wild type
∗
4
∗
∗
∗
∗
∗
∗
∗
2
0
n = 717 509 724 785 256 255 504 482 509 258 250 252 254 251 519 508 528 503 251 251 253 506
Locus of interest
DAPI
0
14D2
11D/E1
14C3
11B5
7F3
10B1
7F5
10B5.3
7F3
15B3.1
7D3
Significant interacting region:
Fetal liver hLCR-AS:
Fetal liver wild type:
Figure 2 Contacts of 8C3–C4 with and without the LCR are similar.
(a) Intrachromosomal DNA interactions of 8C3–C4 with (top) and without
(bottom) an integrated β-globin LCR are essentially similar, as determined
by 4C analysed with a running-mean analysis of microarray data (average
probe spacing: 7 kb). (b) Intrachromosomal DNA interactions of 8C3–C4
with (top) and without (bottom) an integrated β-globin LCR are essentially
similar, as determined by 4C analysed with domainograms that visualize
probability scores (P -values indicated with colour codes) for the clustering
of positive 4C signals over windows ranging in size from 1 to 200 probes.
(c) Interchromosomal 4C data for two chromosomes (7 and 11), analysed
as described above. (d) Validation of 4C results by cryo-FISH; examples of
results. Scale bar: 2 µm. (e) Interaction frequencies with a series of genomic
regions, measured by cryo-FISH in wild-type and LCR transgenic fetal livers.
The number of cells analysed (n) is indicated. Colour codes indicate the
significance of the 4C signal (probe clustering), with green referring to
P < 0.01 and red referring to P ≥ 0.01.
locus on chromosome 7 (Fig. 3a). Homology between the human
and mouse LCRs is limited (<10%) and therefore not expected
to underlie the latter contact. The two loci have in common that
they are very highly expressed and carry genes controlled by EKLF
and GATA-1, two transcription factors that also bind to the LCR.
Ranking genomic regions on the basis of their difference in 4C signal
revealed significant enrichment (P < 0.01, hypergeometric test) of
EKLF (refs 38,39)- and GATA-1 (ref. 40)-regulated genes, as well as
of highly expressed genes, among regions contacted more strongly by
the LCR (Supplementary Table S1). The same conclusion was drawn
946
NATURE CELL BIOLOGY VOLUME 13 | NUMBER 8 | AUGUST 2011
© 2011 Macmillan Publishers Limited. All rights reserved.
ARTICLES
a
Chromosomal location (Mb)
111.0
7E3
110.8
Chromosomal location (Mb)
11A4
32.0
111.2
32.2
1% most highly
expressed genes
0.4
0.2
∗
0
(LCR–Rad23a) bin
0.6
GATA1regulated genes
0.4
0.2
0
∗
< –3
< –2
< –1
>1
>2
>3
>4
>5
>6
>7
0.6
Fraction of probes
0
(LCR–Rad23a) bin
Wild type
hLCR-AS
0.6
EKLFregulated genes
c GATA1
EKLF
28
119
0.4
96
88
0.2
0
∗
7
0
74
< –3
< –2
< –1
>1
>2
>3
>4
>5
>6
>7
2
10
8
6
4
2
0
Fraction of probes
4C signal (k = 21)
α-globin
4
< –3
< –2
< –1
>1
>2
>3
>4
>5
>6
>7
b
Fraction of probes
4C signal (k = 21)
β-globin
32.4
(LCR–Rad23a) bin
1% most highly
expressed
Figure 3 Within a predetermined genomic environment the ectopic LCR
shows preferential interactions with specific genes. (a) 4C results (running
median over sliding windows of 21 probes) for 8C3–C4 with (blue) and
without (red) the integrated LCR, at the endogenous β-globin locus on
chromosome 7 (left) and the α-globin locus on chromosome 11 (right).
(b) Probes were binned according to increasing difference in (LCR −
wild-type) 4C signal and characterized depending on their location relative
to highly expressed and GATA-1- and EKLF-regulated genes. The yellow
dashed line represents the expected frequency on the basis of all probes.
(c) Venn diagram showing the number of, and overlap between, probes
captured more frequently in the integrated LCR 4C experiment for each
category of genes analysed in the population marked with an asterisk in b.
from an analysis per microarray probe (Fig. 3b). Each category of genes
seems to independently attract the LCR (Fig. 3c and Supplementary
Fig. S5b). We conclude that the LCR at 8C3–C4 behaves like a ‘dog
on a lead’: the chromosomal context of 8C3–C4 imposes important
constraints on its freedom to move, but within the restricted nuclear
subcompartment that it occupies the LCR preferentially contacts genes
that are controlled by shared transcription factors.
strategies that measure gene expression across cell populations (Fig. 4d
and Supplementary Fig. S6e).
Interchromosomal gene activation by the LCR
Notwithstanding its restricted ability to change the spatial environment
of 8C3–C4, the ectopic LCR is involved in many long-range interactions
in cis and trans. We investigated whether any of the contacted genes
were upregulated by the LCR. For this, we profiled the transcriptomes
of wild-type and LCR-transgenic littermates using microarrays. A small
number of genes was found to be upregulated more than twofold in
transgenics. These were the previously characterized genes proximal
to the LCR in cis (ref. 18), plus a single gene in trans, Hbb-bh1, that
was confirmed to be upregulated by quantitative PCR with reverse
transcription (RT–qPCR) (Fig. 4 and Supplementary Fig. S6a–d).
Interestingly, Hbb-bh1, or βh1, is one of the endogenous β-globin
genes normally expressed at high, LCR-dependent levels earlier in
development, in embryonic blood cells (Fig. 1a). In E14.5 fetal livers
the gene is looped away from the endogenous LCR (Supplementary
Fig. S1). However, three independent probe-sets on the Affymetrix
microarray and two independent RT–qPCR primer-sets reveal that
the gene is not fully silent, but expressed at basal levels in wild-type
fetal livers (Fig. 4 and Supplementary Fig. S6). Importantly, increased
Hbb-bh1 expression was consistently found in transgenics carrying
the LCR in either one of the orientations when compared with their
wild-type littermates (Fig. 4a,b), but not when 8C3–C4 exclusively
carried a Neo selection cassette (Fig. 4c). No other β- or α-globin
gene detectably changed in expression in the presence of the ectopic
LCR (but see below), as judged by microarray analysis and RT–qPCR
Trans -contacts cause variegated expression
Given that the endogenous β-globin locus carrying βh1 is among the
chromosomal regions that are preferentially contacted by the ectopic
LCR (Fig. 3a), we hypothesized that such interchromosomal LCR–βh1
contacts drive increased βh1 expression. DNA and RNA-FISH analysis
showed that the interaction occurs in 5–10% of the cells, whereas
higher-resolution cryo-FISH identified contacts between 2 and 3%
of the alleles (Supplementary Table S2). Irrespectively, this raises the
question of how these relatively rare interchromosomal contacts can
account for the overall twofold increase in transcript levels measured
across the entire cell population. We reasoned that interaction
frequencies measured by FISH in fixed cells may reflect chromatin
dynamics, such that over time a given interaction occurs in every
cell. Alternatively, they reflect genome conformations that are cell
specific and relatively stable after mitotic exit, implying that, in a
given nucleus, genomic loci sample overlapping or non-overlapping
nuclear subvolumes. If the latter were true, ‘jackpot’ cells should exist
with accumulated βh1 messenger RNA levels in combination with
frequent interchromosomal LCR–βh1 interactions. To investigate this
we carried out RNA FISH using a mixture of probes visualizing both
primary transcripts and accumulated mRNA (ref. 41; Fig. 5a,b). In
wild-type E14.5 fetal liver cells, we failed to detect the βh1 primary
transcript signals that were visible in E10.5 embryonic blood cells
(Supplementary Fig. S7), consistent with its marked drop in expression
during development. Accumulated cytoplasmic βh1 mRNA was seen
in only a small percentage (3.5%) of cells. These cells also contained the
highest adult β-globin transcript levels (Hbb-b1, or β-major) and we
therefore assumed they represent the most differentiated erythroid cells
in the liver that had most time to accumulate β-globin mRNA. Notably,
in only one out of 50 of these cells was an interchromosomal interaction
NATURE CELL BIOLOGY VOLUME 13 | NUMBER 8 | AUGUST 2011
© 2011 Macmillan Publishers Limited. All rights reserved.
947
ARTICLES
a E14.5 fetal liver
Relative
expression
3
Wild type
2
hLCR-AS
1
hLCR-S
0
Rad23 –/–
βh1
βh1
βh1
(1437810_a_at)(1437990_x_at) (1450736_a_at)
b E14.5 fetal liver
c E14.5 fetal liver
Relative
expression
Relative
expression
4
3
2
1
0
Litter:
1
0
βh1
I
βh1
II
βh1
III
βh1
I
βh1
II
βh1
III
βh1
d E14.5 fetal liver
Relative
expression
3
2
1
0
εy
βh1
β maj
β min
ζ-globin
α-globin
Figure 4 The ectopic LCR on chromosome 8 enhances the expression of the
endogenous βh1 gene on chromosome 7. (a) Affymetrix gene-expression
data for all probe-sets analysing βh1 transcripts (n = 3). (b) RT–qPCR
comparison of βh1 gene expression between multiple wild-type and
homozygous LCR-AS littermates, normalized to Hprt1 transcript levels and to
own wild-type littermates. (c) RT–qPCR analysis showing that the insertion
at 8C3–C4 of a neomycin selection cassette instead of the hLCR does not
lead to upregulation of the βh1 gene. Data from two independent samples.
(d) RT–qPCR analysis of expression of β- and α-globin genes in wild-type
and homozygous LCR littermates. Error bars: standard error on the basis of
3 littermates (n).
between 8C3–C4 and β-globin observed, which does not exceed the
interaction frequency of 5.1% measured across the entire wild-type
red-blood-cell population (Fig. 5c and Supplementary Table S2). On
the other hand, in LCR transgenic fetal liver cells the percentage of cells
with high βh1 mRNA levels is increased to 7.0%. More importantly,
in these transgenics 15/50 cells (30%) with high βh1 levels showed
an interchromosomal interaction between 8C3–C4 (with the LCR)
and β-globin, a highly significant (P = 6.6 × 10−7 , hypergeometric
test) increase in interaction frequency when compared with the
8.5% measured across all red blood cells in the transgenics (Fig. 5c
and Supplementary Fig. S8a,b). As a control, we analysed α- versus
β-globin co-localization in transgenic cells and found no correlation
between this interchromosomal interaction and βh1 expression levels
(Supplementary Fig. S8c).
We next reversed the question and asked whether cells showing
the interchromosomal LCR–βh1 interaction also had increased βh1
transcript levels in their cytoplasm as compared with other red blood
cells in the transgenic fetal livers. For this, we developed automated
image-analysis software (see Methods) and analysed thousands of
cells with respect to their intensity of βh1 RNA FISH signal in the
cytoplasm. The analysis showed that in transgenics, but not in wild
type, cells with the interchromosomal interaction between 8C3–C4 and
948
the β-globin locus more often have high βh1 levels than the other cells
from the same tissue (Fig. 6a). Interestingly, albeit less pronounced,
the same was found for β-major (Fig. 6b), the adult β-globin gene
located next to βh1 in the endogenous globin locus (Fig. 1a). Its
natural extremely high expression level precluded uncovering an extra
contribution from the few cells with this interchromosomal interaction
by cell-population-based expression assays such as microarrays or
RT–qPCR. However, our analysis at the single-cell level revealed that
also this adult β-globin gene benefits from contacts with the extra copy
of the LCR. βh1 is looped away from the endogenous mouse LCR and
as such may be available for contacts with the ectopic human LCR.
β-major dynamically forms and breaks contacts with the endogenous
LCR (ref. 7), possibly providing opportunity for the ectopic LCR to also
engage in contacts and further boost transcription. The genes encoding
α-globin did not benefit from contacts with the ectopic LCR, as judged
from the single-cell analysis strategy (Fig. 6c). We conclude that our
transgenic mice contain a unique population of cells that have increased
levels of mRNA for β-globin because these genes on chromosome 7 are
contacted and trans activated by the ectopic LCR on chromosome 8.
DISCUSSION
One of the main challenges of the post-genomic era is to assign
function to genomic sites, many of which have regulatory potential.
Clearly, this cannot be done without considering the dynamics and
spatial configuration of the genome. Here, we uncover properties of
nuclear organization that dictate the action of regulatory elements in
nuclear space. Our findings should contribute to a working model
of genome function. The results demonstrate that regulatory DNA
elements can search for preferred interaction partners, which in the
case of the LCR are genes controlled by shared transcription factors.
The ability to roam the nucleus is however heavily constrained by the
chromosomal context. We predict the same to be true for almost all
genomic locations, although the degree of constraint may vary. The
concept of chromosomal context heavily influencing a gene’s specific
nuclear location seems to contradict more deterministic models of
nuclear organization, where functionally related genes are proposed to
meet at dedicated nuclear sites20 . We cannot exclude that the LCR
would have a more notable effect when placed at other genomic
locations, or that other enhancers exist that are better capable of
repositioning chromosomes and forming specific interchromosomal
interactions. We note however that very few, if any, regulatory elements
have been described with such a strong influence on gene expression
and chromatin organization as the β-globin LCR.
An important finding of this study was that the ectopic, orphan, LCR
on chromosome 8 contacted many different genes in cis and in trans,
including those sharing a similar set of regulatory proteins, but that no
measurable effect on the expression of most of them was detected.
This suggests that, in mammals, ultralong-range gene regulation
within and between chromosomes will be rare, or at least difficult
to measure in cell populations. Two endogenous β-globin genes on
chromosome 7 were the exception, as they were both upregulated
by the ectopic LCR in cells with the relevant interchromosomal
interaction. As these are natural (mouse) target genes of the (human)
LCR, promoter compatibility and spatial proximity seem essential for
transcription regulation over distance. Interestingly, a few examples
exist of endogenous tissue-specific enhancers activating not only target
NATURE CELL BIOLOGY VOLUME 13 | NUMBER 8 | AUGUST 2011
© 2011 Macmillan Publishers Limited. All rights reserved.
ARTICLES
a RNA FISH
c
RNA FISH
E14.5 fetal liver
8C3–C4 versus β-globin locus
b RNA FISH
I
II
III
IV
V
βh1
β maj
8C3–C4
DAPI
Co-localization
frequency (%)
P<< 0.001
30
All
erythroid
cells
20
β h1
'jackpot'
cells
10
NS
0
n = 1,400 50
775 50
Wild type hLCR-AS
Figure 5 Increased βh1 mRNA levels in cells showing interchromosomal
LCR–βh1 interactions. (a) RNA FISH on E14.5 fetal liver cells, with
one cell showing strongly increased βh1 mRNA levels in the cytoplasm
(‘jackpot cell’). Scale bars: 2 µm. (b) Enlargement of the ‘jackpot cell’,
revealing an interchromosomal interaction between the endogenous β-globin
locus on chromosome 7 and the ectopic LCR on chromosome 8. (I–IV)
Probes from one focal plane are shown separately and merged. (V) Z stack
showing all RNA signals for βmaj and 8C3–C4. (c) Quantification of RNA
FISH. Determining interchromosomal interaction frequencies between the
endogenous β-globin locus and 8C3–C4 without (wild type) and with an
integrated LCR (hLCR-AS), in all red blood cells and in ‘βh1 jackpot cells’.
The number of cells analysed (n) is indicated. NS: no significant difference.
genes but also non-target genes that happen to be in physical proximity
to the enhancer42,43 . Our results open the possibility that such bystander
activation may be more common in the genome, but appreciable only
in individual cells that have their genome folded such that an enhancer
and gene happen to be within contacting distance.
Our data provide genetic evidence for classical enhancer activity
between mammalian chromosomes, where the genetic addition or
deletion of a regulatory DNA element on one chromosome causes
increased or reduced expression of a physically interacting gene on
another chromosome. As such, we provide formal in vivo evidence that
mammalian regulatory sites do not need an intervening chromatin
fibre to propagate activating signals to responding gene promoters
elsewhere in the genome, but that spatial proximity, in combination
with enhancer–promoter compatibility, is sufficient for gene activation.
Interchromosomal interactions between mammalian regulatory sites
and genes have been observed before, but genetic evidence for trans
activation was lacking so far. For example, the alternatively expressed
TH 1 and TH 2 cytokine loci were seen to come together before their
activation in naive T cells. On differentiation to T-helper 1 or 2
cells, the interactions between these signature loci were lost and the
respective genes turned on25 . The functional consequences of this
interaction seemed complex, however, and different from classical
enhancer activity. The deletion of a regulatory element in the TH 2
locus caused a delay, rather than a reduction, in the expression of
the TH 1 gene, and intriguingly this effect was measurable only in
differentiating TH 1 cells that no longer showed the interchromosomal
interaction. The interaction was proposed to prepare loci for proper
expression during subsequent T-helper cell specification, an activity
not previously assigned to regulatory sites25 . Interchromosomal gene
regulation by a single enhancer was suggested to control the expression
of all ∼1,200 olfactory receptor genes spread across the genome23 , but
deletion of the enhancer demonstrated that the enhancer only affects
genes in cis44,45 . In another study the activation of human interferon
beta (IFN -β) expression in response to viral infection was reported
to coincide with interchromosomal interactions with three Alu repeat
elements harbouring cryptic NF-κB sites21 . Although transfection
experiments with plasmids carrying these elements supported the idea
that the DNA interactions boost IFN -β expression, formal evidence
for interchromosomal enhancer activity awaits demonstration that
the chromosomal deletion of one of these repeats causes a drop
in IFN -β expression. Finally, the imprinting control region (ICR)
of the H 19-Igf2 locus has been the subject of several studies on
interchromosomal DNA interactions22,24,46 . The data did not reveal
trans activation and were not necessarily consistent, as each study
identified different interchromosomal interactions with different
functional outcomes, possibly owing to the use of different cell types
and/or experimental approaches.
An interesting observation from our artificial system is that
interchromosomal interactions can lead to variegated levels of
accumulated transcripts in the individual cells. We propose to term the
observed phenomenon that cell-specific long-range DNA interactions
cause variable gene expression levels among otherwise identical
cells ‘spatial effect variegation’ (SEV; Fig. 6d). Stochastic cell-to-cell
variation in gene expression, or transcriptional noise, is common in cell
populations47 . Our data open the possibility that SEV may be one of the
underlying mechanisms of transcriptional noise. In such a scenario, the
nature of the interacting region will determine whether gene expression
goes up or down in the corresponding cell. This is different from
position effect variegation48 , where variable expression of ectopically
placed genes is classically thought to be caused by repressive effects
from directly surrounding chromatin. Future research should indicate
if SEV is acting on endogenous genes. If so, it may provide specific cells
within a larger population with a mechanism to make autonomous
cell-fate decisions, without the need for external signalling.
METHODS
Methods and any associated references are available in the online
version of the paper at http://www.nature.com/naturecellbiology
Note: Supplementary Information is available on the Nature Cell Biology website
ACKNOWLEDGEMENTS
We thank J. Marteijn and W. Vermeulen for providing Rad23a knockout material,
Y. Oz for counting FISH slides, V. Buckle for providing BAC probes, J. van Rheenen
and the Hubrecht Imaging Center for help with image analysis and E. Splinter and
other members of the group for assistance. This work was financially supported by
NATURE CELL BIOLOGY VOLUME 13 | NUMBER 8 | AUGUST 2011
© 2011 Macmillan Publishers Limited. All rights reserved.
949
ARTICLES
Pixel count top 1%
1,000
500
100
50
10
5
1
P = 0.0082
0
5,000
0.2
0.4
Quantile rank
0.6
Wild type
Pixel count top 1%
Pixel count top 1%
Pixel count top 1%
hLCR-AS
5,000
1,000
500
100
50
10
5
1
0
β -major—8C3–C4
b
0.2
0.4
Quantile rank
0.6
c
hLCR-AS
5,000
1,000
500
100
50
10
5
1
P = 0.0127
0.2
0.4
0.6
Quantile rank
0
5,000
α-globin—8C3–C4
Pixel count top 1%
β h1—8C3–C4
a
hLCR-AS
5,000
1,000
500
100
50
10
5
1
0
0.2
0.4
Quantile rank
0.6
Wild type
1,000
500
Interacting
100
50
Non-interacting
10
5
1
0
0.2
0.4
Quantile rank
0.6
d Spatial effect variagation:
Gene
Regulatory
element
t=0
t = +1
t = +2
Time
Figure 6 Increased βh1 and β-major mRNA levels in cells showing
interchromosomal LCR–βh1 interactions. (a,b) Automated RNA-FISH
image-analysis (see Methods) results, showing that cells in which
the ectopic LCR interacts in trans with the endogenous β-globin
locus more often have high βh1 (a) or β-major (b) transcript levels
than cells that have the loci apart. (c) Cells in which the ectopic
LCR interacts in trans with the endogenous α-globin locus do not
differ in their levels of mRNA for α-globin from cells without this
interchromosomal interaction. The probability score for the difference
in distributions for interacting and non-interacting cells is calculated by
a one-sided Kolmogorov–Smirnov test. (d) SEV: variegated expression
among otherwise identical cells caused by cell-specific long-range DNA
interactions (intra- or interchromosomal) that are relatively stable during
interphase.
grants from the Dutch Scientific Organization (NWO) (91204082 and 935170621)
and a European Research Council Starting Grant (209700, ‘4C’) to W.d.L.
5.
AUTHOR CONTRIBUTIONS
D.N. and W.d.L. designed the experiments, analysed the data and, with help from
E.d.W., wrote the manuscript. D.N. and P.K. carried out experiments. E.d.W.
analysed 4C data and developed the automated FISH image analysis. H.v.d.W.
analysed 4C and microarray expression data. M.S. carried out 4C experiments. M.L-J.
and R.H.S. designed and synthesized RNA-FISH probes. B.E. and A.d.K. helped with
the FISH experiments.
6.
9.
COMPETING FINANCIAL INTERESTS
The authors declare no competing financial interests.
10.
Published online at http://www.nature.com/naturecellbiology
Reprints and permissions information is available online at http://www.nature.com/
reprints
1.
2.
3.
4.
950
Boyle, A. P. et al. High-resolution mapping and characterization of open chromatin
across the genome. Cell 132, 311–322 (2008).
Amano, T. et al. Chromosomal dynamics at the Shh locus: limb bud-specific
differential regulation of competence and active transcription. Dev. Cell 16,
47–57 (2009).
Kleinjan, D. J. & Coutinho, P. Cis-ruption mechanisms: disruption of cis-regulatory
control as a cause of human genetic disease. Brief Funct. Genomic Proteom. 8,
317–332 (2009).
Epner, E. et al. The β-globin LCR is not necessary for an open chromatin structure or
developmentally regulated transcription of the native mouse β-globin locus. Mol. Cell
2, 447–455 (1998).
7.
8.
11.
12.
13.
14.
15.
16.
Grosveld, F., van Assendelft, G. B., Greaves, D. R. & Kollias, G. Position-independent,
high-level expression of the human β-globin gene in transgenic mice. Cell 51,
975–985 (1987).
Wasylyk, B., Wasylyk, C., Augereau, P. & Chambon, P. The SV40 72 bp repeat
preferentially potentiates transcription starting from proximal natural or substitute
promoter elements. Cell 32, 503–514 (1983).
Wijgerde, M., Grosveld, F. & Fraser, P. Transcription complex stability and chromatin
dynamics in vivo. Nature 377, 209–213 (1995).
Carter, D., Chakalova, L., Osborne, C. S., Dai, Y. F. & Fraser, P. Long-range chromatin
regulatory interactions in vivo. Nat. Genet. 32, 623–626 (2002).
Tolhuis, B., Palstra, R. J., Splinter, E., Grosveld, F. & de Laat, W. Looping and
interaction between hypersensitive sites in the active β-globin locus. Mol. Cell 10,
1453–1465 (2002).
de Laat, W. & Grosveld, F. Spatial organization of gene expression: the active
chromatin hub. Chromosome Res. 11, 447–459 (2003).
Spilianakis, C. G. & Flavell, R. A. Long-range intrachromosomal interactions in the
T helper type 2 cytokine locus. Nat. Immunol. 5, 1017–1027 (2004).
Vernimmen, D., De Gobbi, M., Sloane-Stanley, J. A., Wood, W. G. & Higgs, D. R.
Long-range chromosomal interactions regulate the timing of the transition between
poised and active gene expression. EMBO J. 26, 2041–2051 (2007).
Brown, K. E. et al. Association of transcriptionally silent genes with Ikaros complexes
at centromeric heterochromatin. Cell 91, 845–854 (1997).
Chambeyron, S., Da Silva, N. R., Lawson, K. A. & Bickmore, W. A. Nuclear
re-organisation of the Hoxb complex during mouse embryonic development.
Development 132, 2215–2223 (2005).
Skok, J. A. et al. Nonequivalent nuclear location of immunoglobulin alleles in B
lymphocytes. Nat. Immunol. 2, 848–854 (2001).
Hewitt, S. L. et al. Association between the Igk and Igh immunoglobulin loci
mediated by the 3’ Igk enhancer induces ‘decontraction’ of the Igh locus in pre-B
cells. Nat. Immunol. 9, 396–404 (2008).
NATURE CELL BIOLOGY VOLUME 13 | NUMBER 8 | AUGUST 2011
© 2011 Macmillan Publishers Limited. All rights reserved.
ARTICLES
17. Lundgren, M. et al. Transcription factor dosage affects changes in higher order
chromatin structure associated with activation of a heterochromatic gene. Cell 103,
733–743 (2000).
18. Noordermeer, D. et al. Transcription and chromatin organization of a housekeeping
gene cluster containing an integrated β-globin locus control region. PLoS Genet. 4,
e1000016 (2008).
19. Ragoczy, T., Bender, M. A., Telling, A., Byron, R. & Groudine, M. The locus
control region is required for association of the murine β-globin locus with
engaged transcription factories during erythroid maturation. Genes Dev. 20,
1447–1457 (2006).
20. Schoenfelder, S. et al. Preferential associations between co-regulated genes reveal
a transcriptional interactome in erythroid cells. Nat. Genet. 42, 53–61 (2010).
21. Apostolou, E. & Thanos, D. Virus infection induces NF-κB-dependent interchromosomal associations mediating monoallelic IFN-β gene expression. Cell 134,
85–96 (2008).
22. Ling, J. Q. et al. CTCF mediates interchromosomal colocalization between Igf2/H19
and Wsb1/Nf1. Science 312, 269–272 (2006).
23. Lomvardas, S. et al. Interchromosomal interactions and olfactory receptor choice.
Cell 126, 403–413 (2006).
24. Sandhu, K. S. et al. Nonallelic transvection of multiple imprinted loci is organized
by the H19 imprinting control region during germline development. Genes Dev. 23,
2598–2603 (2009).
25. Spilianakis, C. G., Lalioti, M. D., Town, T., Lee, G. R. & Flavell, R. A.
Interchromosomal associations between alternatively expressed loci. Nature 435,
637–645 (2005).
26. Lewis, E. B. The theory and application of a new method of detecting
chromosomal rearrangements in Drosophila melanogaster. Am. Naturalist 88,
225–239 (1954).
27. Pirrotta, V. Transvection and chromosomal trans-interaction effects. Biochim.
Biophys. Acta 1424, M1-8 (1999).
28. Bolzer, A. et al. Three-dimensional maps of all chromosomes in human male
fibroblast nuclei and prometaphase rosettes. PLoS Biol. 3, 826–842 (2005).
29. Bacher, C. P. et al. Transient colocalization of X-inactivation centres accompanies
the initiation of X inactivation. Nat. Cell Biol. 8, 293–299 (2006).
30. Hewitt, S. L. et al. RAG-1 and ATM coordinate monoallelic recombination
and nuclear positioning of immunoglobulin loci. Nat. Immunol. 10,
655–664 (2009).
31. Xu, N., Donohoe, M. E., Silva, S. S. & Lee, J. T. Evidence that homologous
X-chromosome pairing requires transcription and Ctcf protein. Nat. Genet. 39,
1390–1396 (2007).
32. Branco, M. R. & Pombo, A. Intermingling of chromosome territories in interphase
suggests role in translocations and transcription-dependent associations. PLoS Biol.
4, 780–788 (2006).
33. Strouboulis, J., Dillon, N. & Grosveld, F. Developmental regulation of a complete
70-kb human β-globin locus in transgenic mice. Genes Dev. 6, 1857–1864 (1992).
34. Dekker, J., Rippe, K., Dekker, M. & Kleckner, N. Capturing chromosome
conformation. Science 295, 1306–1311 (2002).
35. Simonis, M. et al. Nuclear organization of active and inactive chromatin domains
uncovered by chromosome conformation capture-on-chip (4C). Nat. Genet. 38,
1348–1354 (2006).
36. de Wit, E., Braunschweig, U., Greil, F., Bussemaker, H. J. & van Steensel, B.
Global chromatin domain organization of the Drosophila genome. PLoS Genet. 4,
e1000045 (2008).
37. Simonis, M., Kooren, J. & de Laat, W. An evaluation of 3C-based methods to capture
DNA interactions. Nat. Methods 4, 895–901 (2007).
38. Drissen, R. et al. The erythroid phenotype of EKLF-null mice: defects in hemoglobin
metabolism and membrane stability. Mol. Cell Biol. 25, 5205–5214 (2005).
39. Hodge, D. et al. A global role for EKLF in definitive and primitive erythropoiesis.
Blood 107, 3359–3370 (2006).
40. Welch, J. J. et al. Global regulation of erythroid gene expression by transcription
factor GATA-1. Blood 104, 3136–3147 (2004).
41. Femino, A. M., Fay, F. S., Fogarty, K. & Singer, R. H. Visualization of single RNA
transcripts in situ. Science 280, 585–590 (1998).
42. Cajiao, I., Zhang, A., Yoo, E. J., Cooke, N. E. & Liebhaber, S. A. Bystander gene
activation by a locus control region. EMBO J. 23, 3854–3863 (2004).
43. Lower, K. M. et al. Adventitious changes in long-range gene expression caused by
polymorphic structural variation and promoter competition. Proc. Natl Acad. Sci.
USA 106, 21771–21776 (2009).
44. Fuss, S. H., Omura, M. & Mombaerts, P. Local and cis effects of the H element on
expression of odorant receptor genes in mouse. Cell 130, 373–384 (2007).
45. Nishizumi, H., Kumasaka, K., Inoue, N., Nakashima, A. & Sakano, H. Deletion of
the core-H region in mice abolishes the expression of three proximal odorant receptor
genes in cis. Proc. Natl. Acad. Sci. USA 104, 20067–20072 (2007).
46. Zhao, Z. et al. Circular chromosome conformation capture (4C) uncovers extensive
networks of epigenetically regulated intra- and interchromosomal interactions. Nat.
Genet. 38, 1341–1347 (2006).
47. Raj, A. & van Oudenaarden, A. Nature, nurture, or chance: stochastic gene expression
and its consequences. Cell 135, 216–226 (2008).
48. Muller, H. J. Types of visible variations induced by X-rays in Drosophila. J. Genet.
22, 299–334 (1930).
NATURE CELL BIOLOGY VOLUME 13 | NUMBER 8 | AUGUST 2011
© 2011 Macmillan Publishers Limited. All rights reserved.
951
METHODS
DOI: 10.1038/ncb2278
METHODS
Gene targeting and generation of transgenic mice. Targeting of the human
β-globin LCR to the mouse Rad23a gene has been described18 . The human Aγ-globin
gene was targeted to the mouse Rad23a gene by substituting the ClaI Neo-resistance
cassette from an existing Rad23a targeting construct by a 7.6 kb ClaI fragment
containing a TK-Neo resistance cassette coupled to a ClaI–SmaI fragment containing
the human Aγ-globin gene with a gene encoding GFP at the translational start.
Constructs with the ClaI fragment in two orientations were obtained: γ-globin-S
(50 -γ-globin at the 50 -end of the Rad23a gene) and γ-globin-AS (50 -γ-globin gene
at the 30 -end of the Rad23a gene). Targeting in Ola129-derived embryonic stem
cells, blastocyst injection and breeding to obtain homozygous transgenic animals
in an FVB background has been described18 . Genotyping was carried out by
Southern blot. Animal experiments were carried out according to institutional
and national guidelines (Committee on Experiments with Laboratory Animals
(DEC-Consult); Ministry of Agriculture, Nature and Food Quality, The Hague,
The Netherlands).
Gene expression analysis. Total RNA was isolated using Trizol (Invitrogen) from
tissues of at least two independent embryos. Complementary DNA was synthesized
using SuperScriptII Reverse Transcriptase and oligo(dT)12–18 primer according to
the manufacturers’ instructions (Invitrogen). Products were quantified by qPCR,
using platinum Taq DNA polymerase (Invitrogen) and SYBR Green (Sigma) on
an Opticon 2 Real-Time PCR detection system (BioRad). Sequences of primers are
provided (Supplementary Table S3). Transcript levels were normalized to the Hprt1
transcript.
FISH. DNA FISH and cryo-FISH was carried out as described before18,35,49 .
Bacterial artificial chromosome (BAC) clones (BACPAC Resources Centre) used to
visualize genomic regions are listed in Supplementary Table S3. For DNA–FISH,
two adjacent BAC probes, RP24-136A15 and RP24-319P23, were labelled with
SpectrumGreen-dUTP (green, Vysis) or ChromaTide Texas Red-12-dUTP (red,
Invitrogen). BAC probes for cryo-FISH were labelled with SpectrumGreen-dUTP
(green) or ChromaTide Alexa Fluor 594-5-dUTP (red; Invitrogen). Probe specificity
was confirmed on mouse spleen metaphase spreads. 500 ng of labelled probe was
co-precipitated with 5 µg of mouse Cot1 DNA (Invitrogen). Images were collected
with a Zeiss Axio Imager Z1 epifluorescence microscope (100× plan apochromat,
1.4 oil objective) equipped with a CCD (charge-coupled device) camera. DNA-FISH
images were analysed with Zeiss AxioVision software (Zeiss). Cryo-FISH images
were analysed with Isis FISH Imaging System software (Metasystems). Filters
used for DNA FISH: DAPI (Zeiss), fluorescein isothiocyanate, AF594 (Chroma).
Filters used for cryo-FISH: DAPI, fluorescein isothiocyanate, RD-TR-PE (Zeiss).
No bleed-through was detected and images were collected without saturation
of intensities. The significance of co-localization (DNA FISH and cryo-FISH)
was determined by applying a replicated goodness-of-fit test (G-statistic). The
null hypothesis in DNA-FISH experiments was that integration of the LCR or
LCR and Aγ-globin gene resulted in similar co-localization frequencies as the
wild-type 8C3–C4 alleles. The null hypothesis in cryo-FISH experiments was that
associating regions identified in the 4C analysis had co-localization frequencies
comparable to those of negative regions. In DNA FISH a minimum of 150
alleles were counted for each genotype; in cryo-FISH a minimum of 500 alleles
were counted by a person not knowing the probe combination applied to
the sections.
To simultaneously detect accumulated mRNA in the cytoplasm and the
nuclear position of transcribed loci, RNA FISH was carried out50 with some
adjustments. E14.5 fetal livers were resuspended in 200 µl PBS and 20 µl was
spotted on a poly-l-lysine slide (Sigma) and air-dried for 5 min at RT before
fixation. For hybridization, slides were rehydrated into PBS plus magnesium and
equilibrated in 50% formamide/2× SSC for 20 min. Hybridization was carried
out three times overnight at 37 ◦ C, followed by a series of post-hybridization
washes and mounting in Prolong Gold antifade reagent (Molecular Probes).
Images were collected on a Leica DM6000 Fluorescence microscope (100 plan
apochromat, 1.4 oil objective), using Leica filters 11513888 (Y5), 11513887 (Y3),
11513890 (GFP) and 11513874 (DAPI), and analysed using Leica Application
Suite software. RNA probes for Hbb-bh1 and Hbb-b1 were synthesized as
described previously50 . A mixture of six different probes was labelled with Cy3
or Cy5 (GE Healthcare). Probe sequences are listed in Supplementary Table S3.
BAC clones RP23-319P23 and 14567 (Incyte), analysing respectively the actively
transcribed locus 8C3–C4 and the active α-globin locus, were labelled with
SpectrumGreen-dUTP (Vysis) as described previously35 . For each combination
of probes (Hbb-bh1 and Hbb-b1 with either 8C3–C4 or α-globin), two to five
slides were independently hybridized and analysed. Images with five to 25 cells
were checked for ‘jackpot cells’, with high levels of βh1 (Hbb-bh1) transcript
in their cytoplasm. After identification of these ‘jackpot’ cells, images for other
probes were opened and cells were analysed for co-localization between the
active endogenous β-globin locus (visualized using Hbb-b1 transcript probes) and
the locus of interest (8C3–C4 or α-globin). Loci were scored as co-localizing
if two spots overlapped or touched in a focal plane without black pixels in
between. Counting results are summarized in Supplementary Table S2. Results were
reproducible between different slides and between different investigators analysing
RNA-FISH slides.
Automated image analysis. For this study, we developed software to automatically
identify and analyse large numbers of microscopy images. In summary: we used the
DAPI and Rad23a signal to define cell boundaries in each image (on average ∼10–20
cells/image). Of all pixels inside cells in an image, we selected pixels with the 1%
highest intensity in the βh1 and β-major signal. Next, we determined the number
of high-intensity pixels for every cell, which is expected to be roughly equal under
the null hypothesis of homogeneous expression.
The R package EBImage (ref. 51) was used for the automated image analysis.
Within a z-stack, the focal point was determined by calculating the absolute
difference between the actual and the blurred version of the image. Because this
is highest in the image with the highest contrast, the focal point is the image
where the sum of this difference is maximal. We binarized Rad23a and DAPI
images by applying a threshold function (thresh) on both channels. This yields two
images with binary information, which are combined into a single binary image
by setting to a value of one pixels that are one in either or both of the original
images (logical OR function). Subsequently, a distance map (distmap) is calculated
using the binary image52 . Cell boundaries were defined by carrying out a watershed
(watershed) function on the distance map53 . 83 wild-type images yielded 1,757 cells;
149 hLCR-AS images yielded 1,563 cells; 214 α-globin images were analysed, yielding
2,108 cells.
Because of the differences in illumination settings between images, a nonparametric threshold rather than an absolute threshold was used for scoring
expression levels. Of all pixels found within cells we selected the 1% with highest
intensity. Homogeneous expression levels would return an equal distribution of
pixels in the cells. However, we see enrichment of high pixels in a population of
high-expressing cells. Cells were then classified as ‘interacting’ (active 8C3–C4 and
β-globin loci overlap or touch), or ‘non-interacting’ (four nascent transcript signals
(two 8C3–C4 and two βmaj ), but no overlap). The top 1% intense pixels in the
βh1 channel (‘intense pixels’) were defined and counted per cell. Per category, cells
were sorted according to their number of intense pixels. Probability scores for the
difference in distributions for interacting and non-interacting cells were calculated
by a one-sided Kolmogorov–Smirnov test.
4C analysis. 4C sample preparation was carried out as described35 , with HindIII
as a primary restriction enzyme, DpnII (wild type) or NlaIII (LCR-AS) as secondary
frequent cutter and EcoNI as tertiary restriction enzyme for linearization. Sequences
of primers are in Supplementary Table S3. Custom Nimblegen arrays carrying
4C probes for chromosomes 7, 8, 10, 11, 12, 14 and 15 were used (release
mm6 of the UCSC build mouse genome; ref. 2). The liftOver tool (UCSC)
was used to transfer probe position to the locations in release mm9 (NCBI37).
Competitive hybridization of 4C product versus genomic DNA was carried out as
described35 . Normalization of the array data was carried out with ‘lowess’, using
standard settings54 . Subsequently, we corrected for differences in ranges between
datasets. Because of the asymmetric, non-normal distribution of the data, standard
variance normalization is not suitable; therefore, we used the median absolute
deviation (a non-parametric estimator for the variance). For the visualization of
DNA interactions we adapted the domainogram algorithm36 , enabling multiscale
visualization of 4C signals. First, 4C data are binarized by calling probes that show
a more than threefold increase of median absolute deviation over the median
positive. Next we take the union (∪) of the positive signals in the two replicates,
which is used as input for subsequent analysis. A probability score for enrichment
of positive 4C signals with a given genomic region is calculated by a binomial
test, and for every chromosome we determine the probability (P) of observing
positive 4C signals by dividing the total number of positive signals by the total
number of probes. For a range of window sizes (w from 2 to 200) enrichment
of positive signals is determined in window w with given P. By carrying out this
analysis using sliding windows, we can analyse entire chromosomes and construct
a multiscale chromosomal map by visualizing the matrix of −log10 P-values of
chromatin interactions.
To identify differentially interacting regions we calculated the average of the two
median absolute deviation normalized replicates followed by the running median
(window size of 21) for both viewpoints for every trans chromosome. Windows with
a difference of two (log 2 scale) were scored as differentially interacting. Genes within
50 kb of a differentially contacted probe were scored as differentially contacted genes.
The resulting genes were intersected with a list of genes that were differentially
regulated on EKLF (refs 38,39) or GATA-1 (ref. 40) deletion, or the 1% most highly
expressed genes. Enrichment for EKLF and/or GATA-1 regulated genes among
NATURE CELL BIOLOGY
© 2011 Macmillan Publishers Limited. All rights reserved.
METHODS
DOI: 10.1038/ncb2278
differentially contacted genes was calculated by comparing this ratio to that of all
EKLF and/or GATA-1 regulated genes among all genes.
Accession numbers. Data are available from Gene Expression Omnibus (GEO):
GSE24614 and GSE5891.
49. Solovei, I. et al. Spatial preservation of nuclear chromatin architecture during
three-dimensional fluorescence in situ hybridization (3D-FISH). Exp. Cell Res. 276,
10–23 (2002).
50. Levsky, J. M., Shenoy, S. M., Pezo, R. C. & Singer, R. H. Single-cell gene expression
profiling. Science 297, 836–840 (2002).
51. Pau, G., Fuchs, F., Sklyar, O., Boutros, M. & Huber, W. EBImage–an R package
for image processing with applications to cellular phenotypes. Bioinformatics 26,
979–981 (2010).
52. Kolountzakis, M. N. & Kutulakos, K. N. Fast computation of the Euclidian distance
maps for binary images. Inform. Process. Lett. 43, 181–184 (1992).
53. Vincent, L. & Soille, P. Watersheds in digital spaces: an efficient algorithm based on
immersion simulations. Pattern Analysis and Machine Intelligence, IEEE Trans. 13,
583–598 (2002).
54. Smyth, G. K. & Speed, T. Normalization of cDNA microarray data. Methods 31,
265–273 (2003).
NATURE CELL BIOLOGY
© 2011 Macmillan Publishers Limited. All rights reserved.
S U P P L E M E N TA R Y I N F O R M AT I O N
DOI: 10.1038/ncb2278
βh1
Figure S1 Spatial organization of the endogenous mouse β-globin locus. Cartoon showing the spatial topology of the mouse endogenous β-globin locus in
definitive red blood cells at E14.5 and later stages of development, as inferred from 3C data. At this stage, βh1 is ignored by its own LCR in the fetal liver.
WWW.NATURE.COM/NATURECELLBIOLOGY
1
© 2011 Macmillan Publishers Limited. All rights reserved.
S U P P L E M E N TA R Y I N F O R M AT I O N
A
Wildtype
H: HindIII
D: DpnII
N: NlaIII
forward: TCACACGCGAAGTAGGCC
reverse: CCTTCCTCCACCATGATGA
H
D
H
HH
H
H
Rad23a
hLCR-AS forward: ACACTTTCAGTCCGGTCC
reverse: AGATTTCCTGTTCACTCACTG
N
H
H
Pu
B
Wildtype
H
hLCR-AS
H
5
4
H H
3
H
H
H
2
HH H
1
C Human β-globin locus
LCR
ε
Gγ Aγ
δ
β
max
4C / control
FL hLCR-AS1
0.0
max
FL hLCR-AS2
0.0
Figure S2 4C analysis of 8C3/C4 with and without the human globin LCR. (a)
Location of restriction fragments and primer sets used for the amplification
of 4C material from wildtype and transgenic animals. Different restriction
fragments and primer sets have been used to analyze DNA interactions
with wildtype 8C3/C4 and 8C3/C4 carrying the integrated human LCR in
either orientation. Only relevant restriction sites of frequent cutters have
been indicated. (b) Gel electrophoresis analysis of PCR amplified wildtype
and transgenic 4C material shows the reproducibility between replicate
samples and the differences between amplified 4C material generated with
different primer sets. (c) Normalized, non-smoothened, 4C signal intensities
for probes mapping to the human δ-globin locus which was present as a
control locus on the dedicated 4C microarray carrying probes for 7 mouse
chromosomes. Positive signal is restricted to the human LCR, while probes
mapping to other parts of the human globin locus are negative.
2
WWW.NATURE.COM/NATURECELLBIOLOGY
© 2011 Macmillan Publishers Limited. All rights reserved.
S U P P L E M E N TA R Y I N F O R M AT I O N
60000
hLCR
chr 8
chr 7
chr 15
40000
chr 14
chr 12
chr 11
20000
chr 10
0
probe count per chromsome
A
1
2
3
4
5
6
1
2
3
4
5
6
1
2
3
4
5
6
60000
40000
20000
0
probe count per chromsome
Wildtype
1.1
1.0
0.9
0.8
0.7
0.6
ratio cis−probes (hLCR/Wildtype)
1.2
B
increasing 4C signal
Figure S3 With the LCR, the ratio of high 4C signal on inter- versus intrachromosomal probes increases. (a) Average normalized 4C signals over two
replicates were sorted in an ascending manner for the wildtype and LCR
experiment. Probes were subsequently divided into six equal-sized bins. Barplots
show for each bin the frequency of probes for every chromosome. Note that in
bin 6 (probes with highest 4C signal) chromosome 8 (the cis chromosome) is
overrepresented, but less so with the LCR than in wildtype. (b) Barplot showing
the ratio of cis probe frequencies between hLCR and WT for the 4C signal
bins. The relative decrease in the amount of cis probes with high intensities is
proportional to more contacts with trans chromosomes by the hLCR.
WWW.NATURE.COM/NATURECELLBIOLOGY
3
© 2011 Macmillan Publishers Limited. All rights reserved.
S U P P L E M E N TA R Y I N F O R M AT I O N
A
4C running mean analysis
Chromosome 8
0
20
40
60
80
100
120
max
hLCR-AS 1
0
max
hLCR-AS 2
0
max
Wildtype 1
0
max
Wildtype 2
0
B
Correlation between 4C replicates:
hLCR
hLCR
Wildtype
hLCR-AS 1
hLCR-AS 2
Wildtype
hLCR-AS 1
hLCR-AS 2
WT 1
WT 2
1
0.64
0.65
0.65
1
0.64
0.59
1
0.77
WT 1
WT 2
Figure S4 Intra-chromosomal contacts of 8C3/C4 without and with LCR
are essentially similar. (a) 4C running mean data in cis of 4 independent
E14.5 fetal liver samples, 2 with and 2 without the ectopic LCR. Despite
1
amplification with different primer sets, the overall chromosome-wide
interaction pattern is comparable. (b) Pairwise correlation coefficients
(Spearman’s rank) between all experiments.
4
WWW.NATURE.COM/NATURECELLBIOLOGY
© 2011 Macmillan Publishers Limited. All rights reserved.
S U P P L E M E N TA R Y I N F O R M AT I O N
4C signal enrichment probability
Wildtype
hLCR-AS
Wildtype
Chromosome 10
Chromosome 14
Chromosomal position (Mb)
Chromosomal position (Mb)
Chromosome 11
Chromosome 15
Window size
Window size
Chromosomal position (Mb)
Window size
hLCR-AS
Chromosomal position (Mb)
Window size
Wildtype
Chromosome 12
Window size
hLCR-AS
Chromosome 7
Window size
A
Chromosomal position (Mb)
Chromosomal position (Mb)
B
GATA1
EKLF
895
All probes:
131
7
119
1053
2
914
GATA1
Probes
preferentially
contacted by
hLCR-AS:
EKLF
28
119
96
88
0
7
total probes:
n=324712
74
1% highest
expressed
1% highest
expressed
Figure S5 The genomic environment of 8C3/C4 without and with LCR is
essentially similar, but the LCR preferentially interacts with genes controlled by
shared transcription factors. (a) Domainograms showing interactions of 8C3/C4
with (top) and without (bottom) the integrated human LCR for all chromosomes
analyzed. Window size: 0-200 probes. (b) Venn diagrams showing the total
number of probes at or around (-/+ 50 kb) the most highly expressed genes,
GATA1-regulated genes and EKLF-regulated genes on the microarray (left) and
among the regions preferentially contacted by the LCR (right).
WWW.NATURE.COM/NATURECELLBIOLOGY
5
© 2011 Macmillan Publishers Limited. All rights reserved.
S U P P L E M E N TA R Y I N F O R M AT I O N
B
Affymetrix gene-expression analysis
Affymetrix gene-expression analysis
E14.5 Fetal liver
4.0
4.0
3.0
3.0
Relative
expression
E14.5 Fetal liver
Relative
expression
A
2.0
1.0
0.0
n=
3
3
3
CalR
3
3
CalR
3
CalR
3
3
3
Syce2
2.0
1.0
0.0
3
3
3
3
Scyl3
(1456170_x_at) (1433806_x_at) (1417606_a_at) (1429270_a_at) (1448717_at)
C
n=
Gcdh
3
3
3
Cdc20 Lsm14a
3
3
βh1
3
3
βh1
3
3
Lactb
(1434365_a_at) (1439394_x_at) (1428437_at) (1437810_a_at) (1437990_x_at) (1449014_at)
D
RT-qPCR analysis
E14.5 Fetal liver
RT-qPCR analysis
E14.5 Fetal liver
primary transcript level
3.0
1.0
0.0
E
2.0
Relative
expression
Relative
expression
2.0
n=
3
3
Scyl3
3
3
4
4
3
Cdc20 Lsm14a
3
4
1.0
0.0
4
Wildtype
hLCR-AS
hLCR-S
βh1
Lactb
βh1
Affymetrix gene-expression analysis
E14.5 Fetal liver
Relative
expression
1.0
0.5
0.0
n=
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
εy
εy
εy
βmaj
ζ-globin
α-globin
α-globin
(1450621_a_at)
(1436823_x_at)
(1436717_x_at)
(1417184_s_at)
(1448716_at)
(1428361_x_at)
(1452757_s_at)
Figure S6 The ectopic LCR on chromosome 8 activates a single gene
in trans. (a) Affymetrix gene-expression data of previously identified
upregulated genes surrounding the integrated LCR. (b) Affymetrix geneexpression data identifying over twofold upregulated genes in trans. (c)
Verification of upregulation in trans by RT-qPCR analysis identifies the
endogenous βh1 gene as the only gene upregulated over twofold in trans.
(d) RT-qPCR of βh1 primary transcript levels demonstrating increased
transcription activity of the βh1 gene at the nascent RNA level. Data
from two independent samples. (e) Affymetrix gene-expression data for
all probe-sets analyzing additional genes in the β- and α-globin loci.
Error bars: Standard Error based on the number of littermates indicated
below (n).
6
WWW.NATURE.COM/NATURECELLBIOLOGY
© 2011 Macmillan Publishers Limited. All rights reserved.
S U P P L E M E N TA R Y I N F O R M AT I O N
A
B
RNA FISH
E10.5 embryonic blood
RNA FISH
E14.5 fetal liver
βh1
βmaj
DAPI
Figure S7 RNA FISH analysis with probes specific for βh1 and βmaj transcripts. In E10.5 embryonic blood, βh1 primary transcripts and mRNA is detected,
but βmaj transcripts not. Scale bars: 2 μm
WWW.NATURE.COM/NATURECELLBIOLOGY
7
© 2011 Macmillan Publishers Limited. All rights reserved.
S U P P L E M E N TA R Y I N F O R M AT I O N
A
RNA FISH
E14.5 FL
1
B
C
2
zoom ins:
1
Z-stack:
1
RNA-FISH
E14.5 Fetal liver
Co-localization
frequency (%)
30
α-globin locus vs.
β-globin locus
zoom ins:
2
Z-stack:
2
βh1
8C3/C4
βmaj
DAPI
20
10
0
NS
n=574
n=44
hLCR-AS
Figure S8 RNA-FISH showing increased βh1 mRNA levels in cells with
inter-chromosomal LCR-βh1 interactions. (a) RNA-FISH on E14.5 fetal
liver cells, with one cell in both fields showing strongly increased βh1
mRNA levels in the cytoplasm (‘jackpot cell’). (b) Zoom in on the ‘jackpot
cell’ highlighted in a., showing an inter-chromosomal interaction between
the endogenous β-globin locus on chromosome 7 and the ectopic LCR on
chromosome 8. The zoom-in shows separate images of three channels from
a single focal plane and in the bottom right a image of the merged βmaj and
8C3/C4 is shown. The z-projection shows the entire z-stack highlighting
all primary transcript-signals for βmaj and 8C3/C4. Scale bars: 2 μm (c)
No increased βh1 transcript levels in cells showing an interchromosomal
interaction between the β- and α-globin locus. Quantification of interchromosomal interactions in LCR-AS transgenics between the β-globin locus
on chromosome 7 and the α-globin locus on chromosome 11, in e14.5
fetal liver total erythroid cells and in ‘βh1 jackpot cells’. Number of cells
analyzed (n) is indicated.
8
WWW.NATURE.COM/NATURECELLBIOLOGY
© 2011 Macmillan Publishers Limited. All rights reserved.
S U P P L E M E N TA R Y I N F O R M AT I O N
Supplementary Tables
Table S1 Genomic regions that are differentially contacted when the LCR is integrated in 8C3/C4. Genomic regions are ranked based on their difference in 4C
signal. The presence of EKLF and GATA-1 regulated genes in these regions is indicated.
Table S2 FISH results. (a) Interaction frequencies measured in E14.5 fetal liver cells as measured by RNA-FISH. (b) Interaction frequencies measured
in E14.5 fetal liver cells, between the endogenous β-globin locus on chromosome 7 and 8C3/C4 on chromosome 8, without (WT) and with (hLCR-AS) the
integrated hLCR. Measurements were made by cryo-FISH (which necessarily focuses on alleles), DNA –FISH (where interactions were scored only in cells
showing signals for all 4 alleles) and RNA-FISH (where interactions were scored only in the 50-80% of red blood cells in the fetal liver (i.e. cells showing
cytoplasmic β-major transcripts). Note that cryo-FISH gives much lower interaction frequencies, as expected from a technique providing ~5 -fold higher
resolution. Note also that frequencies measured by different techniques are difficult to compare, as they all score interactions differently.
Table S3 List of primers and BACs.
WWW.NATURE.COM/NATURECELLBIOLOGY
9
© 2011 Macmillan Publishers Limited. All rights reserved.
A. RNA-FISH results Analyzed loci 8C3/C4 vs β-globin 8C3/C4 vs β-globin α-globin vs β-globin Tissue Total β-globin
expressing
cells hLCR-AS Wildtype hLCR-AS 775 1400 574 Colocalization in βglobin expressing
cells
cells % 66 71 15 8.5 % 5.1 % 2.6 % βh1 positive cells
Colocalization in
(among β-globin
expressing cells)
cells % βh1 positive cells
50 50 44 cells % 15 1 1 30.0 % 2.0 % 2.2 % 6.5 % 3.6 % 6.7 % B. Comparison of FISH results (interaction frequency 8C3/C4 – β-globin locus)
FISH technique Tissue Cryo-FISH hLCR-AS Wildtype hLCR-AS Wildtype hLCR-AS Wildtype DNA-FISH
RNA-FISH
Total cells 534 262 129 90
775
1400
counted 8C3/C4 alleles Cells with DNAFISH signals Cells with βmaj
mRNA signal
Colocalization n
%
12 6 8 8
66
71
2.3 % 2.3 % 6.2 % 8.9 %
8.5 %
5.1 %
© 2011 Macmillan Publishers Limited. All rights reserved.
Significance of difference
(Fisher exact test)
Not significant Not significant p = 0.00437
4C primers
Primer set
Sequence
Rad23a
TCACACGCGAAGTAGGCC
CCTTCCTCCACCATGATGA
ACACTTTCAGTCCGGTCC
AGATTTCCTGTTCACTCACTG
hLCR-AS
Gene ID and Affymetrix probes
Gene name
Gene ID (ENSEMBL)
Rad23a
CalR
ENSMUSG00000003813
ENSMUSG00000003814
Syce2
ENSMUSG00000003824
Scyl3
Cdc20
Lsm14a
βh1 (Hbb-bh1)
ENSMUSG00000026584
ENSMUSG00000006398
ENSMUSG00000066568
ENSMUSG00000052217
Lactb
εy (Hbb-y)
ENSMUSG00000032370
ENSMUSG00000052187
βmaj (Hbb-b1)
βmin (Hbb-b2)
ζ-globin (Hba-x)
α-globin (Hba-a1)
ENSMUSG00000073940
ENSMUSG00000052305
ENSMUSG00000055609
ENSMUSG00000069919
Affymetrix Probes
1417606_a_at
1433806_x_at
1456170_x_at
1429270_a_at
1448717_at
1434365_a_at
1439394_x_at
1428437_at
1437810_a_at
1437990_x_at
1450736_a_at
1449014_at
1436823_x_at
1436717_x_at
1450621_a_at
1417184_s_at
1448716_at
1428361_x_at
1452757_s_at
Gene expression analysis
Primer set
Sequence
mRNA Hprt
AGCCTAAGATGAGCGCAAGT
ATGGCCACAGGACTAGAACA
TTCACCACCATGGAGAAGGC
GGCATGGACTGTGGTCATGA
AGGTGCTGACTTCCTTGGG
GGGTGAATTCTTTGCCGAA
TGGACAACCTCAAGGAGAC
AGTAGAAAGGACAATCACCAAC
GAACTTGTCCTCTGCCTCT
ATCACCAGCACATTACCCA
ATGCCAAAGTGAAGGCCCAT
CCCAGCACAATCACGATCAT
ATCCCAAGGTGAAGGCCCAT
CCCAGCACAATCACGATCGC
GAGAGAGCTATCATCATGTCC
AAGTAGGTCTTCGTCTGGG
TGGCCATGGTGCTGAATATG
TCTTGCCGTGACCCTTGAC
TCTGGGAGTTGAGACTGTGA
TGGACCCATGGACTCTAACA
CTTACCATCTGGACTTGCTG
GGGTGACGGAGTGTCTTTA
TGCTCCATCCTCTGGTCT
CGTGCTGTGTGTCCTTTG
CCACCCAAACCACAATGT
GGACTGAACTGACTGTATGC
CGTGGTTGGAGTTTCTGTAG
TGCTGATGCTTGCGATTC
mRNA Gapdh
mRNA human Aγ-globin
mRNA βh1
mRNA εy
mRNA βmaj
mRNA βmin
mRNA ζ-globin
mRNA α-globin
Primary transcripts βh1
mRNA Scyl3
mRNA Cdc20
mRNA Lsm14a
mRNA Lactb
© 2011 Macmillan Publishers Limited. All rights reserved.
BAC clones for DNA-FISH and Cryo-FISH
BAC clone
RP23-317H16
RP23-370E12
RP23-32C19
RP23-265I23
RP23-27B18
RP24-136A15
RP24-319P23
RP23-87K3
RP24-130O14
RP23-375D18
14567 (Incyte)
RP24-236L11
RP23-311P1
RP23-258M10
RP23-450E9
RP24-255K10
Locus
(Cytological band)
7D3, negative control
7E3, β-globin locus
7F3, negative control
7F3
7F5
8C3 (previously 8C3/C4)
8C3 (previously 8C3/C4)
10B1
10B5.3, negative control
11A4, α-globin locus
11A4, α-globin locus
11B5
11D/E1
14C3
14D2
15B3.1
Chromosome
Position (Mb)
(NCBI assembly m37)
7
7
7
7
7
8
8
10
10
11
11
11
11
14
14
15
93
111
131
134
148
87
87
41
74
32
32
78
102
56
70
37
RNA-FISH probes
Probe
Sequence
βh1 5UTR
GTGAGGTCTAGAAGCTTGGAGATGATCTCAAGTGTGCAAAAGCCAGAATG
βh1 Exon 1
TCCAAGTCCACTTTATCCCAGATGCTTGTGATAGCTGCCTTCTCCTCAGC
βh1 Intron 1
TGTGCCACAAAACCCTATAGAAACCCTGGAAATTTCTGCCATGCATAAGG
βh1 Intron 2
TCAAACCCAAGGCCCCAGAAATGCTGGGCGCTCACTCAAATCTGCACCCA
Alternate βh1 3'UTR 1
GGGCATCATAGACACATGGGATTGCCAGTGTACTGGAATGGAGTTTAATG
Alternate βh1 3'UTR 2
TTAAACAAAATTTGTGCTCTCAATGCCTAATCCAGTCCCCATGGACTCAA
βmaj 5’-UTR1
CTGTTTCTGGGGTTGTGAGTCAACACAACTATGTCAGAAGCAAATGTGAG
βmaj Exon1
TTCATCGGAGTTCACCTTTCCCCACAGGCAAGAGACAGCAGCCTTCTCAG
βmaj Intron1
GGAAAGCAGACCTCTGTCTCCAAGCACCCAACTTCTTCTTGTGAGCTGCC
βmaj Intron2
ACTCCACACACAGTCATGGAGACTGCTCCCTAGAATCGCTTCCCCTTTTT
Alternate bmaj 3’-UTR1
CTTGGGAACAATTAACCATTGTTCACAGGCAAGAGCAGGAAAGGGGGTTT
Alternate bmaj 3’-UTR2
AGAAGACAGATTTTCAAATGTCTATCATTTTGCCAACAACTGACAGATGC
© 2011 Macmillan Publishers Limited. All rights reserved.